Complex Wavelet-Based Sinogram Segmentation for Metal Artifact Reduction in Cone-Beam CT

TL;DR

This work proposes a projection-domain metal artifact reduction method based on analytical metal segmentation in the three-dimensional sinogram using the three-dimensional Dual-Tree Complex Wavelet Transform, where directional wavelet coefficients are exploited to extract the wavefront set and singular support of metal structures.

Abstract

Metal objects pose a significant challenge in cone-beam computed tomography, as their strong and energy-dependent X-ray attenuation leads to inconsistent projections and severe streaking and shading artifacts in reconstructed images. These artifacts degrade image quality and limit the reliability of subsequent medical analysis. We propose a projection-domain metal artifact reduction method based on analytical metal segmentation in the three-dimensional sinogram using the three-dimensional Dual-Tree Complex Wavelet Transform, where directional wavelet coefficients are exploited to extract the wavefront set and singular support of metal structures. The resulting segmentation enables projection-domain inpainting and artifact-reduced reconstruction by combining metal-free and metal-only reconstructions. The proposed approach is evaluated on both simulated and clinical cone-beam computed tomography data and consistently reduces metal artifacts compared to conventional image-domain hard-thresholding methods. The results demonstrate improved visual quality and robustness in clinically realistic scenarios, highlighting the potential of analytically grounded, non-learned projection-domain segmentation for metal artifact reduction.

Figures (4)

• Figure 1: Sinogram (left) and the sum of the absolute value of the largest complex wavelet coefficients, revealing the location of the metals (right).
• Figure 2: The proposed workflow for metal artifact reduction, based on projection-domain metal segmentation. Steps include: A. Extracting the wavefront set related to metals from the 3D sinogram. B. Binary mask for metal segmentation. C. Sinogram inpainting and metal-free reconstruction. D. Metal reconstruction. E. Final result: combining the artifact-free inpainted reconstruction and metal reconstruction.
• Figure 3: Metal segmentation results for simulated cases (I) and (II), presenting $xy$- and $xz$-slices of the 3D CBCT sinograms. Note that the metal mask slightly overshoots the metal area over the neighboring slice due to the morphological closing operation.
• Figure 4: Comparison of FDK reconstructions before correction and after artifact reduction using HT-MAR (comparison method) and CW-MAR (proposed method). Results are shown for physical phantoms with simulated metals (I–II) and experimental metals (III–V). For each phantom and method, axial, sagittal, and coronal slices are displayed. Reconstructions are not HU-calibrated. Visualization uses dataset-specific HU-equivalent windows to highlight metal-induced artifacts. Windowing (range, center): (I) $[-1000, +1400]$ HU (200 HU), (II) $[-2700, +5700]$ HU (1500 HU), (III) $[-996, +186]$ HU ($-405$ HU), (IV) $[-998, -310]$ HU ($-654$ HU), and (V) $[-1240, -30]$ HU ($-635$ HU).